AI Update 2026: Autonomous AI Agents, AI Security Risks, AI Search & the Future of Work | Ep 280 | DevReady Podcast

·43 min·2 clips

27,000 malicious skills and sleeper agents pose threats that overshadow any autonomous agent benefits.

Anthony opens by treating the current agent boom as a lot of hype that still needs deflating. He and Gareth Ryden from Froyo treat the new tools as assistants that follow prompts, not as AGI, and not as a secret new world quietly taking shape behind the scenes. The office joke lands because it feels plausible and absurd at once: picture an open-plan workplace with 50 people talking to agents at the same time, then try to imagine the noise and overlap. Gareth is back after about six weeks. He says there has been enough new AI news to make the last stretch feel crowded, and the conversation settles into that sense of a field moving fast. Hype does not last long here. When the talk turns to coding, Gareth describes a workflow that starts with a product brief, then a detailed PRD, then epics and stories, and only after that a new Claude Code project loaded with markdown files. He is still learning vibe coding. But he likes the structure because it gives the model room to help instead of room to guess, and he says that if you do not give it enough information, it will make assumptions and never tell you. The method is plainspoken. He would rather have the model present a few approaches, explain the differences, and then let him pick one than try to stack multiple tools at once and hope they cooperate. That takes time. Gareth says he now spends about 80 percent of his effort planning, 10 percent building, and 10 percent debugging, which makes the usual software grind feel more deliberate and less random. The payoff is easy to see. He likes finally seeing the thing instead of staring into a black screen, and the episode keeps coming back to that old developer pleasure of watching an idea turn into a tool. The mood stays skeptical. Anthony keeps pressing on shallow AI claims, and Gareth keeps dragging the discussion back to what actually happens inside a working team, a real office, or a real build process. The closer gets more playful. They end by daydreaming about ski goggles that show live performance data, wind and snow conditions, and micro-adjustments, which leaves the conversation on a neat note about AI surfacing useful information in the moment.

As heard by us

A skeptical, practical look at agent hype and a planning-first coding workflow.

Anthony and Gareth spend the episode trimming away agent hype and returning to the practical question of what these tools actually do once they enter the workplace.

Read the full review in PlayNext →

Why you'd press play

If you want the agent hype stripped back, this stays practical.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from DevReady Podcast